Connect Spring 1998  Statistics and Social Sciences

A New Testing Technique
Maintaining Anonymity for Student Participants

Frank LoPresti

Researchers who conduct studies in which participants are tested before and after a treatment have a legal and ethical need to preserve the participants' anonymity. How can we ask people to participate in a two-test process without compromising their privacy? Of course, we want to match the initial test to the follow-up test to get the strongest statistical power, so at some point we must know the names of the respondents.

Most researchers handle this data collection problem by numbering the tests -- giving identical numbers to the pre-test and post-test documents of each participant -- and keeping some "master list" with names matched to questionnaire I.D. numbers. The master list is kept secure by the researcher in a separate location from the pre-tests. After the post-test is administered to the group, the list is destroyed, the matching tests are paired and the respondents remain anonymous.

I was recently involved in testing several thousand children in the NYC public school system who were participating in a violence prevention effort called "The Leadership Program." The in-school program, conducted by a not-for-profit community agency, would span two months.

The Privacy Problem

Since special program money is tight, school boards are demanding more accountability. Testing is becoming more and more important. Nevertheless, the budget was small. The tests were to be administered in the classroom. Twenty different schools were involved. Developing a master list would clearly be a difficult and expensive task, because people must be hired to be responsible for the lists, who would also travel to the schools and administer the tests. The trainers, unfortunately, were a diverse group of part-time actors. They were good with the children, but were difficult to manage from a central location since they resisted administrative tasks.

We could not involve the teachers in keeping a secure ID/names list, as the students' privacy might be compromised.

In the previous year, to avoid creating a master list, the program designers tried an innovative idea. They asked the children for shoe size, number of siblings, age, gender and other information which, when combined, might make matching possible. Responses like "My sneaker size is 6 but my dress shoe size is 5," and "I have two brothers, and sometimes two half-sisters," lead to a match rate of about one in ten, with lots of uncertainty on the accuracy of the matches.

Our New Method of Administering Tests

Pairs of questionnaires with the same IDs were handed out. Each child answered the pre-test and wrote his or her name on an envelope containing the post-test. The pre-tests were collected by the trainers and returned to the program office. The as-yet unanswered post-tests, in their envelopes, were kept at the schools.

Two months later, the envelopes were distributed to the children, who threw away the envelopes with their names on them and filled out the questionnaires inside. The school authorities were satisfied that privacy was maintained, and the matching rate was wonderful.

I welcome any feedback readers may have on this new method. Please e-mail comments to frank.lopresti@nyu.edu. [ C ]

Posted January 20, 1998